Abstract <p>This study presents a research analytics platform for large-scale materials science literature analysis, leveraging over 110,000 abstracts from MRS conferences (2011–2024). We introduce a novel two-stage hierarchical clustering methodology combining LLM-based semantic keyword extraction (approx 550,000 keywords) with optimized embedding-based clustering, resolving memory scalability constraints while preserving semantic coherence. The resulting Streamlit platform enables multi-granular analysis across three clustering levels and nine analytical pathways, including temporal trend analysis, global research distribution across 195+ countries, and AI-powered semantic search. This work provides researchers with unprecedented tools for intuitive visualization of materials science trends and worldwide research distributions.</p> Graphical abstract <p></p>